Interpreting the formation of co-author networks via utility analysis

Interpreting the formation of co-author networks is an interesting task since it can uncover the human behaviour reasons why the co-author network can form, and also a challenging one because the evolution process of co-author network cannot be observed from time to time. Based on a single observation of a co-author network, this paper aims to uncover the inherent reasons of forming the co-author links via adopting utility analysis and estimating preference parameters embedded in the designed utility function. Not only does the developed method consider the author's utility change before and after forming a link, but it also takes into consideration the latent meeting sequence that is crucially important to determine the network formation. In order to estimate these preference parameters accurately and effectively, a double Markov chain Monte Carlo (MCMC for short) algorithm is further developed with low time complexity and capacity of inferring the evolution process with latent variables. Besides, the performance of the proposed method is further assessed under different meeting probabilities and various evolution periods in a series of simulated networks via numerical experiments, and an application is also designed to demonstrate how the method works in a real-world co-author network. Results shows that the proposed method remains accurate and robust in parameter estimation under different settings of synthetic networks and outperforms other selected approaches in predicting the status of network formation. In all, the proposed analytic method can be regarded as a new tool in uncovering the behaviour reasons of co-author network formation in particular and be potentially applied in a wide range of networks in general.